#11266: in_tile_headers not reset in jpeg2000_read_main_headers()
-------------------------------------+-------------------------------------
Reporter: Pierre- | Type: defect
Anthony Lemieux |
Status: new | Priority: critical
Component: avcodec | Version: git-
| master
Keywords: j2k | Blocked By:
Blocking: | Reproduced by developer: 0
Analyzed by developer: 0 |
-------------------------------------+-------------------------------------
in_tile_headers is a context variable that is preserved across calls to
jpeg2000_read_main_headers(), and thus across frames in a multi-frame
file. Once set to 1 it remains equal to 1 forever, even though we might
not be in a tile header.
This is an issue since some processing in jpeg2000_read_main_headers()
depends on the value of in_tile_headers. In particular, checking whether
COD markers are present tile part headers, which is prohibited for
homogeneous HT codestreams.
To make a long story short, processing an MXF file that contains HT
codestreams will fail on the second frame.
{{{
./ffmpeg -i
mer_shrt_23976_vdm_sdr_rec709_g24_3840x2160_20170913_12bit_DCDM.4bpp.mxf
out.mp4
}}}
throws
{{{
[jpeg2000 @ 0x55fac474ee00] COD marker found in a tile header but the
codestream belongs to the HOMOGENEOUS set
[vist#0:0/jpeg2000 @ 0x55fac47400c0] [dec:jpeg2000 @ 0x55fac474e380] Error
submitting packet to decoder: Invalid data found when processing input
}}}
NOTE: It is not clear why in_tile_headers is a context variable since it
looks like it is only used in jpeg2000_read_main_headers().
Sample file at:
https://www.sandflow.com/public/mer_shrt_23976_vdm_sdr_rec709_g24_3840x2160_20170913_12bit_DCDM.4bpp.mxf
--
Ticket URL: <https://trac.ffmpeg.org/ticket/11266>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker_______________________________________________
FFmpeg-trac mailing list
FFmpeg-trac@avcodec.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-trac
To unsubscribe, visit link above, or email
ffmpeg-trac-requ...@ffmpeg.org with subject "unsubscribe".
